| *** eggmaster has quit IRC | 01:14 | |
| *** coolsvap has joined #openstack-rpm-packaging | 02:56 | |
| *** pcaruana has joined #openstack-rpm-packaging | 07:03 | |
| *** jpena|off is now known as jpena | 07:43 | |
| *** makowals has quit IRC | 08:12 | |
| *** makowals has joined #openstack-rpm-packaging | 08:13 | |
| *** itxaka has joined #openstack-rpm-packaging | 08:36 | |
| *** amoralej_ has joined #openstack-rpm-packaging | 08:53 | |
| *** pcaruana has quit IRC | 08:54 | |
| *** dteselkin has quit IRC | 08:54 | |
| *** amoralej has quit IRC | 08:54 | |
| *** dteselkin has joined #openstack-rpm-packaging | 08:56 | |
| *** pcaruana has joined #openstack-rpm-packaging | 09:00 | |
| toabctl | IgorYozhikov, astsmtl any comment on the proposed config schema handing from https://review.openstack.org/#/c/379784/ ? | 09:10 |
|---|---|---|
| toabctl | or jpena ? | 09:10 |
| *** apevec has joined #openstack-rpm-packaging | 09:10 | |
| *** apevec has joined #openstack-rpm-packaging | 09:10 | |
| toabctl | apevec, any comment on the proposed config schema handing from https://review.openstack.org/#/c/379784/ ? | 09:10 |
| jpena | toabctl, I like the idea, just like number80 commented this is quite similar to what we do in RDO. The main difference is that we also have another config file with distro defauls | 09:18 |
| apevec | we'll need to abstract Version, 7.0.0~b3 is not following Fedora guideline | 09:19 |
| apevec | I'll add inline comments | 09:19 |
| toabctl | jpena, I thought the upstream config is still in /etc/$service/$service.conf in your setup. no? | 09:20 |
| apevec | jpena, I guess we could try to finally push that patch to oslo.config for dist.conf | 09:20 |
| apevec | toabctl, it is, overriding distro defaults in /usr/share/$service/$service-dist.conf | 09:20 |
| apevec | order of config files matters, latter override former | 09:21 |
| toabctl | apevec, jpena but do you ship a upstream config template somewhere? | 09:23 |
| *** Guest10079 is now known as zigo | 09:23 | |
| jpena | toabctl, yes, at /etc/$service/$service.conf | 09:24 |
| apevec | yes, as generated in /etc | 09:24 |
| apevec | ah I see suse has different approach | 09:24 |
| apevec | /etc is empty | 09:24 |
| toabctl | so /etc/$service/$service.conf overrides /usr/share/$service/$service-dist.conf ? | 09:24 |
| toabctl | apevec, it's a proposal. we don't use that yet | 09:25 |
| apevec | the unmodified config provided by upstream - that's all commented out right? | 09:25 |
| apevec | as effectively == empty | 09:25 |
| apevec | (as generated by o-c-generator ) ? | 09:25 |
| toabctl | apevec, yes | 09:25 |
| toabctl | apevec, but it's still useful to know all the config options | 09:25 |
| apevec | so no point in shipping that in usr/share which users are not supposed to modify | 09:25 |
| apevec | toabctl, that woudl be /usr/shar/doc/ then | 09:26 |
| toabctl | apevec, ca I download somewhere the openstack-neutron rpm for Newton to have a look? | 09:26 |
| apevec | you can grab latest trunk builds from http://buildlogs.centos.org/centos/7/cloud/x86_64/rdo-trunk-master-tested/ | 09:27 |
| apevec | ah newton | 09:27 |
| apevec | then http://mirror.centos.org/centos/7/cloud/x86_64/openstack-newton/ | 09:27 |
| apevec | or from CBS Koji http://cbs.centos.org/koji/packageinfo?packageID=283 | 09:28 |
| toabctl | apevec, /usr/share/doc would also work for that. | 09:28 |
| apevec | here is dist.conf for neutron https://github.com/rdo-packages/neutron-distgit/blob/rpm-master/neutron-dist.conf | 09:28 |
| toabctl | apevec, so you are not shipping a sample config, right? | 09:29 |
| apevec | it is shipped in /etc/ | 09:29 |
| apevec | but then config mgmt can overrwrite it | 09:30 |
| apevec | so yes, we're missing unmodified copy for doc purposes | 09:30 |
| apevec | we should add that to /usr/share/doc/$service | 09:30 |
| number80 | *nods* | 09:31 |
| toabctl | apevec, but then an empty /etc/$service/$service.conf would be fine, right? | 09:31 |
| apevec | effectively yes | 09:31 |
| apevec | unless someone uses things like augeas and uncommenting options :) | 09:32 |
| * apevec assumes not | 09:32 | |
| toabctl | apevec, that would still work I think | 09:32 |
| apevec | puppet is just using ini writer | 09:32 |
| apevec | afaik | 09:32 |
| number80 | I prefer having working config files for people not using cfgmgmt | 09:33 |
| toabctl | number80, an empty file would still work | 09:33 |
| apevec | toabctl, actually, yeah, augeas lens would need to know about all options, and I don't think anyone has done it for openstack configs? | 09:33 |
| number80 | toabctl: service would not be running | 09:33 |
| apevec | number80, we ship all comented out conf | 09:33 |
| number80 | I mean you can launch it but it would be plain unusable | 09:34 |
| apevec | as generated by o-c-gen | 09:34 |
| apevec | number80, that;s why you need deployment tool | 09:34 |
| toabctl | number80, you would have the dist.conf and and empty /etc/$service/$service.conf . that should be enough. no? | 09:34 |
| apevec | it is not | 09:34 |
| number80 | empty or commented out? | 09:34 |
| toabctl | apevec, why not? | 09:34 |
| apevec | none of openstack RPMs work out of the box | 09:34 |
| apevec | w/o configuring at least db | 09:35 |
| apevec | and creating db schema before that | 09:35 |
| toabctl | apevec, with "should work" I mean it's the same as it is today | 09:35 |
| apevec | s/before/after/ | 09:35 |
| apevec | toabctl, ok then | 09:35 |
| toabctl | number80, is there a difference between empty and all commented? | 09:35 |
| apevec | I thought you need it would work out of the box :) | 09:35 |
| number80 | toabctl: for end-users, yes | 09:35 |
| apevec | there is n't | 09:35 |
| toabctl | number80, add a README to /etc/$service which explains the config structure | 09:36 |
| apevec | number80, but they read docs in /usr/share/doc/ don't they ? | 09:36 |
| apevec | number80, is there anyone doing manual openstack config by reading config files? | 09:36 |
| toabctl | hopefully not :-) | 09:36 |
| apevec | they at least start with upstream install guide | 09:36 |
| number80 | apevec: not always, I'd rather not find thread on m-l about us shipping broken packages | 09:36 |
| number80 | last time, it was because we had missing keys in the config fie | 09:37 |
| apevec | it would be the same, but yeah, I think we can do both: | 09:37 |
| apevec | all commented out in /etc | 09:37 |
| apevec | and copy in /usr/share/doc/ | 09:37 |
| number80 | *nods* | 09:38 |
| toabctl | but /etc/$service/$service.conf is not overridden by a package update. so you would end up with an old template after an update | 09:38 |
| toabctl | a release-update I mean | 09:38 |
| apevec | if not touched, it would be over written on rpm update | 09:38 |
| toabctl | well, true | 09:38 |
| apevec | and if user touches it, they're on their own and we respect their right to shut themselves into whatever body part they like | 09:39 |
| toabctl | sure | 09:39 |
| toabctl | I still like an empty /etc/$service/$servic.conf more because after an update there is no missleading "doc" in that file (because I think nobody does use an unchanged $service.conf) | 09:40 |
| toabctl | also configuration management systems usually override /etc/$service/$service.conf so users shouldn't modifiy that anyway. that's why I would like to have the /etc/$service/conf.d/common dir | 09:41 |
| apevec | I don't think mixing conf mgmt + user modification is a good idea, and we'll definitely want input from puppet, ansible etc. folks on config layout proposal | 09:45 |
| toabctl | apevec, there are situations where you config management system doesn't support featureX but you still want to configure it | 09:46 |
| apevec | ah interesting use-case | 09:47 |
| toabctl | apevec, for the available systems, nothing would change. they just use /etc/$service/$service.conf . | 09:47 |
| toabctl | apevec, also if you add e.g. a 3rd party plugin (e.g. for Neutron) which is not supported by your config management system. that's imo not unusual | 09:48 |
| number80 | yep, and you prefer such changes to be in a separate file | 09:48 |
| apevec | ok, for neutron plugins use-case we'll need to include our networking folks, current setup is all Ihar's | 09:49 |
| number80 | https://www.youtube.com/watch?v=JAmZhXtIIXc <= Ihar's talk about configuration | 09:51 |
| *** amoralej_ is now known as amoralej | 10:00 | |
| apevec | from https://youtu.be/JAmZhXtIIXc?t=4m33s is explanation of autodiscovery and then rdo patch for /usr/share dist.conf | 10:01 |
| toabctl | apevec, is that -dist.conf patch already proposed upstream? | 10:03 |
| apevec | I don't think, it was done long ago by pbrady | 10:04 |
| apevec | we should propose it and see how it goes | 10:04 |
| toabctl | but I anyway like more the explicit way to add --config-dir/--config-file parameters when launching the service. that's easier to debug imo | 10:05 |
| toabctl | I think I'll write a mail to openstack-dev about the config handling topic. | 10:06 |
| number80 | ack | 10:08 |
| apevec | yeah, thinking about it, better move to explicit --config param vs pushing that oslo patch upstream - but IIRC it was also to have those defaults from CLI, I'll put task in RDO to review this | 10:19 |
| toabctl | apevec, number80 mail sent. | 11:03 |
| number80 | thanks :) | 11:12 |
| openstackgerrit | Merged openstack/pymod2pkg: Fix k8sclient to k8s https://review.openstack.org/383036 | 11:29 |
| *** amoralej is now known as amoralej|lunch | 12:06 | |
| *** jpena is now known as jpena|lunch | 12:33 | |
| *** amoralej|lunch is now known as amoralej | 13:15 | |
| *** coolsvap has quit IRC | 13:16 | |
| *** jpena|lunch is now known as jpena | 13:52 | |
| *** apevec has quit IRC | 14:24 | |
| *** openstack has joined #openstack-rpm-packaging | 18:24 | |
| *** openstack has joined #openstack-rpm-packaging | 19:05 | |
| number80 | dirk, toabctl, IgorYozhikov, jpena|off: https://review.openstack.org/#/c/384662/ | 19:10 |
| number80 | it's still early for the main repo but that's also something we should change at some point (likely release:trailing) | 19:11 |
| dirk | number80: yeah, saw that comment, thanks for doing the change request | 19:16 |
| openstackgerrit | Dirk Mueller proposed openstack/rpm-packaging: Update keystoneclient to 3.6.0 https://review.openstack.org/384668 | 19:23 |
| openstackgerrit | Dirk Mueller proposed openstack/rpm-packaging: Update keystonemiddleware to 4.10.0 https://review.openstack.org/384669 | 19:23 |
| openstackgerrit | Dirk Mueller proposed openstack/rpm-packaging: Update keystoneauth1 to 2.13.0 https://review.openstack.org/384670 | 19:23 |
| openstackgerrit | Dirk Mueller proposed openstack/rpm-packaging: Update python-openstackclient to 3.3.0 https://review.openstack.org/384671 | 19:23 |
| openstackgerrit | Dirk Mueller proposed openstack/rpm-packaging: Update osc-lib to 1.2.0 https://review.openstack.org/384672 | 19:23 |
| openstackgerrit | Merged openstack/renderspec: Remove filter for license https://review.openstack.org/384472 | 23:26 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!